Systems architected for the frontier,
built to endure and lead.

The operational backbone behind every product in this portfolio: tiered model routing across Opus, Sonnet, and Haiku; role-specialized executive and engineering agents; fail-closed CI security-review gates verified adversarially with bait pull requests; cross-machine agent federation; and skill supply-chain allowlisting. I designed and packaged it as a versioned, installable plugin, so the same engineering discipline, not just the code, travels with every venture I build.

The public RAYFINITE site, plus the internal platform that runs the company behind it: identity and access control, HR, goals and approvals, asset tracking, and an append-only audit trail. Both are live and still growing, and the intranet is where the operating rules the rest of the portfolio follows are actually enforced.

• Identity & Access
• HRMS
• Approvals
• Audit Trail
• Firebase

A SwiftUI + HealthKit iOS app that correlates weather patterns with personal health metrics, paired with a companion web experience on Firebase/GCP. I designed and built it end to end as a consumer wellbeing product, the same architecture discipline as the enterprise systems above, applied to a lighter, personal surface.

HonestResolve turns empty resolutions into real accountability across iOS, Android, and web, under the Suncalp brand. I architected a 5-step AI-guided "Honest Interview" that scores commitment authenticity (Bronze/Silver/Gold badges), streak-based check-ins, and social accountability features, a wall, leaderboards, reactions, and direct messaging, with Sign in with Apple and a Firebase-backed architecture.

LovelyRobo is a spatial AI-companion platform: photoreal 3D companion robots, each a full task-specialist (career coaching, productivity, and more) that completes real workflows end to end rather than just chatting, with live voice conversation, persistent memory, and warm character. I architected it to be offered both individually and as a bundle.

A multi-tenant community platform connecting diaspora communities with local businesses, cultural events, jobs, and discussion forums, where one codebase serves several separately branded communities. I architected the tenant isolation, the data model, and a search layer that runs embeddings locally and answers questions from retrieved listings and forum posts rather than from the model alone.

• Multi-Tenant
• Next.js
• Postgres + pgvector
• RAG
• Community

Zenolon is RAYFINITE's ERP intelligence platform, four core modules (Governance/Identity, HRMS/HCM, Financials, and the Zenolon Engine) that other RAYFINITE products integrate with, rather than are owned by. The Zenolon Engine module is a CRDT-based distributed coordination store, letting AI agents on multiple machines share state, tasks, and knowledge without a central server. I architected it as the integration layer underneath FactGuard, Dataveb, AEGIDON, AThEnA, and DiamondHull: designed once, reused across the portfolio.

FactGuard is the first product out of the Zenolon Suite: an evidence verification engine live across web, a public REST API, and an MCP server on Firebase/GCP, with the iOS app built and preparing for launch. verification engine with multi-provider model routing and a circuit breaker for resilience, plus live Stripe subscription billing, end to end, from evidence grounding to production deploy.

A self-serve, no-code web-scraper and automated-reporting SaaS built on Python/FastAPI, Next.js, and Playwright. I architected it security-first: a point-and-click selector-builder UI, an SSRF-guarded extraction layer, an idempotent job scheduler, and Stripe billing with signed-webhook verification and per-tenant usage metering.

DiamondHull is the security layer across the portfolio, built as a portable, deploy-anywhere purple-team platform for home and small-office networks: passive-first intrusion detection that never sits in the path of a legitimate device, decoys and honeypots, verifiable evidence capture, and multi-channel alerting, with a red-team module that validates the detections. Response is manual by default, so a false positive never cuts off a real device. The evidence layer signs each bundle through a swappable asymmetric signing port, specified for Ed25519, so an auditor can verify a bundle without holding a shared secret.

A cloud/edge hybrid multi-agent platform for IT incident response, built on FastAPI, WebSockets, PostgreSQL, and Redis. I architected an incident-commander orchestration model that coordinates six specialist agents under a graduated permission system, zero-trust credential handling, blast-radius analysis before any action is taken, and end-to-end audit logging throughout.

AThEnA is the platform for making consequential decisions better: AI-augmented decision support for regulated and sovereign environments, where generic tools cannot go. It integrates model-agnostic inference, mission alignment, risk analysis, decision workflow, and audit into one system, so a decision can be traced, explained, and defended to a regulator rather than merely produced. Fourteen functional domains, specified in Corporate and Government editions and designed to deploy across cloud, on-premises, and classified environments. It presents nothing it cannot trace to a source, separates what is verified from what is inferred, and says so plainly when it does not know. It is the largest specification effort in the portfolio: business requirements, product requirements, a functional requirements document spanning twenty sections, a design-patterns catalog, and a conceptual model, all locked before a line of the build.